Lecture
Enabling intelligent Mg-sheet processing utilizing efficient ML-algorithm
Magnesium sheets for their comparably light weight and physical properties have a high potential to change traditionally die casting issues concerning endurance, strength, ductility, etc. One of the most favored processes to produce magnesium sheets, which are competitive in cost and properties, is the production of thin strips by twin roll casting [1]. The two key parameters of this process are temperature (T) and degree of deformation (φ). After production, in order to analyze the texture of sheets a formal way is the exposure of x-ray, synchrotron or neutron and study the diffraction data. The MATLAB™ toolbox MTEX is very helpful in this regard by offering the so called “pole figure” to study crystallographic preferred orientation [2]. In this contribution we proposed a machine learning algorithm, which replicates an input-output relation between T and φ as input and a mathematical basis for pole figure. We showed that with this algorithm the texture analysis step (exposure and pole figure computation) can be skipped after the training is performed. With sufficient training the algorithm predicts a set of manufacturing process parameters lead to what type of texture. This contribution is limited to pole figure based texture analysis, however we work on the extensions for other options in future. This is an important step to pave the way for industrial and application based manufacturing of Magnesium sheets.
| Category | Short file description | File description | File Size | ||
|---|---|---|---|---|---|
| Extended Abstract | Version 1 | This is a short manuscript to the abstract. | 46 KB | Download |